As Mike Magee demonstrates , a little roadmap can tell you an awful lot.
Item #1: No Prescott until a year from now. That little roadmap now says 4Q. Even if Intel subsequently push it up a couple months, we’re still talking early fall for it.
Problems with manufacturing .09 micron chips? Maybe, but I think paying to build .09 micron fabs is the more likely cause for the slowdown.
Intel has been less-than-aggressive in its pricing moves lately. They haven’t been acting like a company out to do AMD in, more like a company looking first towards its bottom line.
The stock market has been less than impressed with Intel’s “invest through the recession” strategy, especially as the U.S. computer depression continues. AMD isn’t going to be able to trump them throughout most of the year, anyway. There’s some signs Intel has figured out how to get a little more out of the Northwood core than initially expected.
So slow down the fab changes a bit to make Wall Street happy, get one more processor out of Northwood, speed things up a little bit with higher buses, and use hyperthreading to sell it all. That would appear to be the Intel strategy.
Item #2: Hyperthreading slashed in price, but . . . Intel will introduce hyperthreading at lower speeds by late February. They will charge a premium for them. However, that premium will be as low as $15 for the 2.4GHz.
Charging $500 extra dollars for hyperthreading and hyping the hell out of it is one thing, charging $15 extra for it is quite another.
However, there may be a hidden cost to pay, and we overclockers may face a Hobson’s choice.
Item #3: Intel to migrate quickly to 200MHz bus HT chips Intel will get rid of the 400MHz FSB chips and introduce 800MHz FSB chips that will include HT. That’s clear.
However, the article seems to imply that only 800MHz FSB chips (outside of the 3.06) will have HT. We can’t expect any more big increases in memory speed until DDR II.
That means overclockers will have a choice. Unless dual DDR provides so much bandwidth that it wouldn’t be a bottleneck to run FSB faster than memory speed (and the boards available let you do that), you can overclock, or you have HT. Not both.
Analysis: Intel will confine its aggression to its chipsets for most of 2003 (jumping to dual-channel DDR400 is very aggressive, may too much so). The rest will be a big song-and-dance for hyperthreading.
This is excellent news for AMD, better than any they’ve had for quite some time, better than anything that can come from Apple. It means Intel can’t strangle even a delayed Hammer with .09 micron chips.
It also leaves Barton in not-too-bad shape, either; people can upgrade to that and not be too far behind in 2003.
It means AMD won’t be too far behind Intel in developing .09 micron CPUs (provided they can keep to their roadmap).
It makes next spring a deciding point between a number of options. By, say, April, people should have a very good idea what a Barton and Hammer and hyped-up PIV platform will get them, and decide accordingly.